[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"pack-detail-mcp-monitoring-logs-fr":3,"seo:pack:mcp-monitoring-logs:fr":92},{"code":4,"message":5,"data":6},200,"操作成功",{"pack":7},{"slug":8,"icon":9,"tone":10,"status":11,"status_label":12,"title":13,"description":14,"items":15,"install_cmd":91},"mcp-monitoring-logs","📊","#0891B2","stable","Stable","MCP Monitoring + Logs","Neuf serveurs MCP et sous-agents qui transforment Prometheus \u002F Grafana \u002F Sentry \u002F Datadog \u002F SigNoz en quelque chose qu'un agent IA peut vraiment interroger. Logs, métriques, traces, dashboards, alertes — câblés pour que 'pourquoi checkout a explosé à 02:47 ?' obtienne une réponse sans qu'un humain ne fasse grep.",[16,28,35,42,50,57,67,77,84],{"id":17,"uuid":18,"slug":19,"title":20,"description":21,"author_name":22,"view_count":23,"vote_count":24,"lang_type":25,"type":26,"type_label":27},3608,"818380f9-674d-5217-88ab-f393ff99a247","signoz-mcp-server-query-traces-logs-alerts","SigNoz MCP Server — Query Traces, Logs & Alerts","SigNoz MCP Server connects MCP clients to your SigNoz instance: query traces\u002Flogs, inspect alerts, and automate observability workflows using an API key.","MCP Hub",261,0,"en","mcp","MCP",{"id":29,"uuid":30,"slug":31,"title":32,"description":33,"author_name":22,"view_count":34,"vote_count":24,"lang_type":25,"type":26,"type_label":27},827,"655bef8a-41bb-4eda-8f5b-29d1d4cb8c74","axiom-mcp-log-search-analytics-ai-agents-655bef8a","Axiom MCP — Log Search and Analytics for AI Agents","MCP server that gives AI agents access to Axiom log analytics. Query logs, traces, and metrics through natural language for AI-powered observability and incident response.",272,{"id":36,"uuid":37,"slug":38,"title":39,"description":40,"author_name":22,"view_count":41,"vote_count":24,"lang_type":25,"type":26,"type_label":27},3191,"0be32c8b-2ad9-47f2-aa64-26f9a7f6f2c3","grafana-mcp-dashboards-alerts-oncall-tools","Grafana MCP — Dashboards, Alerts & OnCall Tools","Grafana MCP server connects your MCP client to Grafana so agents can search dashboards, query panels, and automate ops tasks with a service account token.",46,{"id":43,"uuid":44,"slug":45,"title":46,"description":47,"author_name":48,"view_count":49,"vote_count":24,"lang_type":25,"type":26,"type_label":27},2853,"b309576b-0970-46f2-b0f8-a2f1af76eeb1","datadog-mcp-server-query-metrics-and-logs-from-ai-agents","Datadog MCP Server — Query Metrics and Logs from AI Agents","Community Datadog MCP server lets Claude \u002F Cursor query metrics, logs, traces, monitors in natural language. SRE copilots, on-call triage.","Datadog",212,{"id":51,"uuid":52,"slug":53,"title":54,"description":55,"author_name":22,"view_count":56,"vote_count":24,"lang_type":25,"type":26,"type_label":27},665,"a739e813-e8fa-4285-8634-55aa447dd71a","sentry-mcp-error-monitoring-server-ai-agents-a739e813","Sentry MCP — Error Monitoring Server for AI Agents","MCP server that connects AI agents to Sentry for real-time error monitoring. Query issues, analyze stack traces, track regressions, and resolve bugs with full crash context. 2,000+ stars.",320,{"id":58,"uuid":59,"slug":60,"title":61,"description":62,"author_name":63,"view_count":64,"vote_count":24,"lang_type":25,"type":65,"type_label":66},2276,"676b8063-2e21-49ce-89eb-020bcc40cb47","sentry-errors-auto-triage-subagent-for-sentry-676b8063","sentry-errors — Auto-Triage Subagent for Sentry","Open-source Claude Code subagent that pulls recent Sentry errors via MCP, groups by component, and suggests fix priorities. Inspired by Boris Cherny.","Skill Factory",239,"skill","Skill",{"id":68,"uuid":69,"slug":70,"title":71,"description":72,"author_name":73,"view_count":74,"vote_count":24,"lang_type":25,"type":75,"type_label":76},3546,"2c1c7883-fd56-5cb3-91fe-6f9d953193f3","pup-datadog-cli-companion-for-ai-agents","Pup — Datadog CLI Companion for AI Agents","Pup is an Apache-2.0 Datadog CLI with 200+ commands across 33+ products, so agents can query logs, metrics, RUM, and security data via OAuth login.","Script Depot",150,"script","Script",{"id":78,"uuid":79,"slug":80,"title":81,"description":82,"author_name":22,"view_count":83,"vote_count":24,"lang_type":25,"type":26,"type_label":27},3444,"71f97e34-fa9c-5c0b-8c21-69d6570cb21f","langfuse-mcp-query-langfuse-traces-via-mcp","langfuse-mcp — Query Langfuse Traces via MCP","Connect Langfuse observability to Claude Code\u002FCodex via MCP: fetch traces, prompts, and datasets (37 tools). Works with Langfuse Cloud or self-hosted.",231,{"id":85,"uuid":86,"slug":87,"title":88,"description":89,"author_name":73,"view_count":90,"vote_count":24,"lang_type":25,"type":65,"type_label":66},3335,"a86f3430-eb78-50ab-bebe-6eef4f53ea4a","monoscope-llm-query-for-logs-traces-metrics","Monoscope — LLM Query for Logs\u002FTraces\u002FMetrics","Monoscope stores logs\u002Ftraces\u002Fmetrics in S3-compatible buckets and lets you explore them with natural-language queries plus a CLI and self-hosted UI.",177,"tokrepo install pack\u002Fmcp-monitoring-logs",{"pageType":93,"pageKey":8,"locale":25,"title":94,"metaDescription":95,"h1":96,"tldr":97,"bodyMarkdown":98,"faq":99,"schema":115,"internalLinks":120,"citations":133,"wordCount":146,"generatedAt":147},"pack","MCP Monitoring + Logs — 9 Agent-Facing MCPs for Prometheus, Grafana, Sentry & Datadog","SigNoz MCP, Axiom MCP, Grafana MCP, Datadog MCP, Sentry MCP, sentry-errors triage subagent, Pup (Datadog CLI), langfuse-mcp, Monoscope — wire your AI agent into the monitoring stack you already run. Logs → metrics → traces → alerts → dashboards, in install order.","MCP Monitoring + Logs — The Agent-Facing Layer of Your Observability Stack","Nine MCPs in deliberate install order: log query MCP first (SigNoz, Axiom), then metrics + dashboard MCPs (Grafana, Datadog), then error \u002F alert MCPs (Sentry + sentry-errors triage subagent), then CLI companion (Pup), then LLM trace MCPs (langfuse-mcp, Monoscope). The goal: an agent reads Sentry, pulls the SigNoz trace, queries Loki via MCP, and writes a one-paragraph incident summary — without you opening Grafana.","## What this pack is — and isn't\n\nThis pack is not a monitoring stack. **You already have one.** Prometheus is scraping, Grafana is rendering, Sentry is grouping exceptions, maybe Loki or SigNoz is eating logs. The pack is the *agent-facing layer* on top of that stack — the MCP servers and subagents that let Claude, Cursor, ChatGPT, or any MCP-aware client query traces, logs, alerts, and dashboards conversationally instead of through a human clicking around.\n\nIf you want the underlying log pipeline (winston, Fluent Bit, Loki, ClickHouse, lnav), the [log-analysis-search pack](\u002Fen\u002Ftopics) covers that. If you want the broader deploy + monitor stack (Vercel, Uptime Kuma, OpenTelemetry Collector, Alertmanager), the [deploy-monitor-observability pack](\u002Fen\u002Ftopics) covers that. This pack assumes those exist and adds the MCP layer that lets an agent actually use them.\n\nWhy this matters: most teams already pay for or self-host enough observability tooling. What they don't have is an on-call workflow where an alert auto-triages itself, an SRE asking \"what changed at 02:47?\" gets a real answer in one sentence, and a deploy-time incident gets a written summary before anyone opens a dashboard. That's what MCP unlocks — if you wire it deliberately.\n\nEvery pick here is **open-source or has an open-source server you self-host**. Several wrap commercial backends (Datadog, Sentry SaaS, Axiom) but the MCP server itself is open code you read before running. No black-box agent SDKs.\n\n## Install in this order\n\n1. **SigNoz MCP Server** — log + trace + alert query MCP. Start here because SigNoz is one of the few open-source backends that natively unifies logs, traces, and metrics in one store. Once an agent can query SigNoz, it can answer \"slowest endpoint last hour,\" \"first occurrence of this error,\" and \"alerts firing now\" without three separate tools. If you're not on SigNoz, skip to step 2.\n2. **Axiom MCP — Log Search and Analytics for AI Agents** — cloud log search MCP. Axiom is the alternative when your logs live in a hosted store with APL (Axiom Processing Language) instead of LogQL. Same job as SigNoz MCP, different backend. Pick whichever matches where your logs already are; running both is overkill.\n3. **Grafana MCP — Dashboards, Alerts & OnCall Tools** — dashboard MCP. This is the keystone. Grafana's MCP exposes panel data, alert rules, OnCall schedules, and dashboard search. Once installed, an agent can pull the same chart you'd open manually, read the underlying PromQL\u002FLogQL, and reason about it. Without this, the agent is blind to everything Grafana is rendering.\n4. **Datadog MCP Server — Query Metrics and Logs from AI Agents** — commercial alt for shops on Datadog. Same role as steps 1+3 combined, but against Datadog's metrics + logs + APM. Read-only is the default safety posture — verify before exposing to an autonomous agent.\n5. **Sentry MCP — Error Monitoring Server for AI Agents** — error MCP. Sentry's MCP is the official server; it returns issue lists, stack traces, regression status, and release health. An agent triaging an alert at 3 a.m. starts here — \"is this a new issue or a known regression?\" is a one-tool answer.\n6. **sentry-errors — Auto-Triage Subagent for Sentry** — alerting workflow agent. This is the layer above Sentry MCP: a Claude subagent that wakes when an alert fires, calls Sentry MCP to fetch the issue, calls SigNoz\u002FGrafana MCP for context, and posts a structured triage note. You can write your own, but this one's already proven; fork it.\n7. **Pup — Datadog CLI Companion for AI Agents** — CLI bridge. When the agent's MCP toolset can't express a query (or you don't trust MCP exposure for write ops), Pup gives the agent a sandboxed Datadog CLI. Read-only flags are the default; pair with audit logging.\n8. **langfuse-mcp — Query Langfuse Traces via MCP** — LLM-trace MCP. Different category: instead of querying your app's request traces, this queries the *LLM* traces in Langfuse — prompt, response, tool calls, cost. When the on-call question is \"why did the agent answer wrong at 02:47?\" instead of \"why did checkout fail?\", this is the tool. Mandatory if you ship LLM features to prod.\n9. **Monoscope — LLM Query for Logs\u002FTraces\u002FMetrics** — unified natural-language query layer. Sits in front of multiple backends (logs, traces, metrics) and lets an agent or human ask \"show me errors from checkout in the last 30 minutes with p99 > 500ms\" without picking a tool first. The right pick when you've outgrown asking each MCP individually.\n\n## How they fit together\n\n```\n   [ alert fires ]\n         │\n         ▼\n   sentry-errors subagent           ← wakes up, orchestrates\n         │\n         ├──▶ Sentry MCP             (issue, stack trace, regression?)\n         ├──▶ SigNoz \u002F Axiom MCP     (logs around the timestamp)\n         ├──▶ Grafana MCP            (panel state, alert rule, OnCall roster)\n         ├──▶ Datadog MCP \u002F Pup CLI  (metrics if that's where you are)\n         ├──▶ langfuse-mcp           (LLM trace, if AI feature was in path)\n         │\n         ▼\n   Monoscope (optional)              ← unified NL query, fan-out\n         │\n         ▼\n   [ structured triage note → ticket \u002F Slack \u002F paging ]\n```\n\nThe shape is deliberate: **subagent on top, MCP servers below, backend stores at the bottom**. Each MCP is a thin shim; the value is composition. A single SigNoz MCP call is interesting; sentry-errors fanning out across 4 MCPs and writing a paragraph is the actual unlock.\n\nIf you're starting from zero: skip steps 4 (Datadog) and 7 (Pup) unless you're on Datadog. Skip step 8 (langfuse-mcp) if you don't ship LLM features. The minimal viable rig is **SigNoz MCP + Grafana MCP + Sentry MCP + sentry-errors subagent** — four picks, agent can answer most 3 a.m. questions.\n\n## Tradeoffs you'll hit\n\n- **MCP vs custom function-calling** — every MCP server here could be re-implemented as OpenAI function-calling against the same backend API. MCP wins when you have more than one agent runtime (Claude Desktop, Cursor, ChatGPT custom GPTs, internal agents) — write the MCP once, every client uses it. Function-calling wins for one bespoke agent with one client.\n- **Read-only vs read-write MCP scope** — every monitoring MCP here can be configured read-write (acknowledge alerts, silence alarms, create dashboards). For on-call triage, read-only is the only sane default. Read-write is a separate decision per server, with audit logging mandatory.\n- **Open-source MCP vs vendor MCP** — Grafana MCP and Sentry MCP are first-party from Grafana Labs and Sentry. SigNoz MCP and Axiom MCP are also first-party. Datadog MCP and Pup are community-maintained against Datadog's API. First-party is more stable; community moves faster on edge features. Read the maintainer before deploying.\n- **sentry-errors vs roll-your-own subagent** — sentry-errors is an opinionated triage flow. If your incident playbook is different (you page first, triage later) it'll feel wrong. Fork it; the value is the fan-out pattern, not the exact prompts.\n- **Monoscope vs individual MCP servers** — Monoscope is a unifier. You don't need it until you have 4+ monitoring MCPs and an agent that's spending too many tokens picking which one to call. Start without it; add when fan-out latency or token cost becomes a real problem.\n\n## Common pitfalls\n\n- **Exposing write scopes by default** — every MCP server doc shows the write example first because it's flashier. For monitoring\u002Fobservability MCPs specifically, the agent should not be able to silence alerts, ack incidents, or modify dashboards without an explicit human-approved path. Audit every MCP config before deploy.\n- **Agent token cost on large log queries** — an agent asking \"all errors in the last 24h\" via SigNoz MCP can pull megabytes into context. Cap response sizes in the MCP server config; reject queries above N rows and tell the agent to add filters.\n- **Mixing logs and metrics in one MCP call** — the agent will try. Most backends answer one well at a time. Encode the discipline in the system prompt: ask SigNoz MCP for the trace ID first, then ask Grafana MCP for the metric panel, not both in one query.\n- **No correlation between MCP servers** — `trace_id` is the glue. If your logs, traces, and metrics don't share a `trace_id`, the agent's fan-out will fetch four unrelated things and hallucinate a connection. Fix instrumentation before fixing the agent.\n- **MCP server running on the same VM as production** — read-only MCP still consumes memory + CPU under burst agent usage. Run MCPs on a separate small VM; isolate from your observability backend so a runaway agent can't OOM your log store.\n- **sentry-errors firing without rate limit** — if alerts storm (deploy regression, infra event), the subagent will wake on every one. Add a deduplicate window at the alert source (Alertmanager \u002F Sentry rules) or rate-limit the subagent's invocations; don't pay Anthropic API for 500 triage notes about the same root cause.",[100,103,106,109,112],{"q":101,"a":102},"Do I need all nine of these, or is there a minimum viable rig?","Minimum is four: SigNoz MCP (or Axiom MCP if your logs live there) + Grafana MCP + Sentry MCP + sentry-errors triage subagent. That covers logs, dashboards\u002Falerts, errors, and the orchestration layer. Add Datadog MCP and Pup only if your shop is on Datadog. Add langfuse-mcp only if you ship LLM features and need LLM trace visibility. Add Monoscope only when fan-out across 4+ MCPs costs too many tokens. Most teams land at five to six picks.",{"q":104,"a":105},"Why is this pack different from log-analysis-search or deploy-monitor-observability?","Those packs are the underlying observability stacks: log-analysis-search covers winston\u002FLoguru\u002FFluent Bit\u002FLoki\u002FClickHouse\u002Flnav (how logs get stored and read), deploy-monitor-observability covers Prometheus\u002FGrafana\u002FUptime Kuma\u002FAlertmanager (how the stack gets deployed). This pack is the layer on top that lets an AI agent query the stacks those packs build. There's deliberate complementarity — install one of those first, then this one.",{"q":107,"a":108},"Should MCP servers be read-only or read-write?","Read-only by default for everything in this pack. The job is observation, not mutation. Read-write MCPs (ack alerts, silence alarms, modify dashboards) are a separate decision per server, require audit logging, and should go through a human-approval step in the agent workflow. The risk asymmetry — agent reading dashboards vs agent silencing a real alert — makes read-only the unambiguous default.",{"q":110,"a":111},"What's the difference between an MCP server and an MCP subagent like sentry-errors?","An MCP server (Sentry MCP, SigNoz MCP, Grafana MCP) exposes one backend's API as a set of tools an agent can call. An MCP subagent (sentry-errors) is a higher-level agent that calls multiple MCP servers in sequence to accomplish a workflow — receive alert → fetch issue → fetch logs → fetch dashboard panel → write triage note. Servers are the primitives; subagents are the composed workflows. You usually run both: servers as infrastructure, subagents as automation.",{"q":113,"a":114},"Can this pack work with my existing on-call tools (PagerDuty, Opsgenie, custom)?","Yes, with a shim. The sentry-errors subagent and Grafana OnCall integration assume an alert source — once an alert reaches the agent (webhook, Sentry rule, Alertmanager receiver), the fan-out across MCPs is alert-source-agnostic. PagerDuty and Opsgenie both expose webhooks; route them into the same agent endpoint. The pack doesn't replace your paging tool, it triages before the human gets paged.",{"@context":116,"@type":117,"name":13,"description":118,"numberOfItems":119,"inLanguage":25},"https:\u002F\u002Fschema.org","ItemList","Nine MCP servers and subagents in install order: log query MCPs (SigNoz, Axiom), dashboard MCPs (Grafana, Datadog), error MCPs (Sentry + auto-triage subagent), CLI bridge (Pup), and LLM-trace MCPs (langfuse-mcp, Monoscope) — the agent-facing layer of your monitoring stack.",9,[121,125,129],{"url":122,"anchor":123,"reason":124},"\u002Fen\u002Fai-tools-for\u002Fobservability","AI tools for observability","These MCPs are the agent-facing entry point into the broader observability catalog",{"url":126,"anchor":127,"reason":128},"\u002Fen\u002Ftopics","Browse other topic packs","Pair with the log-analysis-search and deploy-monitor-observability packs for the full stack",{"url":130,"anchor":131,"reason":132},"\u002Fen\u002Ffeatured","Featured assets on TokRepo","These nine MCPs sit alongside the broader curated catalog",[134,138,142],{"claim":135,"source_name":136,"source_url":137},"Model Context Protocol is an open standard for connecting tools to LLM agents","Model Context Protocol specification","https:\u002F\u002Fmodelcontextprotocol.io\u002F",{"claim":139,"source_name":140,"source_url":141},"Grafana provides a first-party MCP server exposing dashboards, alerts, and OnCall tools","Grafana MCP documentation","https:\u002F\u002Fgrafana.com\u002Fdocs\u002Fgrafana\u002Flatest\u002F",{"claim":143,"source_name":144,"source_url":145},"Sentry provides an official MCP server for AI agents to query errors and issues","Sentry MCP server","https:\u002F\u002Fdocs.sentry.io\u002F",920,"2026-05-22T00:00:00Z"]